Most reviews/articles I have seen of water cooling systems focus on achieving lowest possible delta-T from ambient, with 10C or below as the golden target and anything above 15C is an “overloaded” water cooling system. This review is different because I am going to focus on whether a single 120x120 radiator is sufficient to handle the heat from a powerful contemporary graphics card like the NVidia GTX580. This exercise was borne out of my curiosity on whether a “low capacity” radiator can dissipate enough heat from a powerful card going at full blast while at the same time remaining within the quietness standards of SPCR readers. On my side is Newton’s law of cooling that states the rate of heat loss is proportional to the differences in temperature between the bodies, or in our case, the difference in temperatures between the ambient air and the radiator. One trip to the store and my wallet fifty bucks lighter yielded the following review.
The GTX580 is NVidia’s latest top-of-the-line graphics card. With an official TDP of 250W, and some claims on the web of reaching up to 350W, it’s no slouch when it comes to generating heat. NVidia got a lot of heat (pun intended) from enthusiasts from the noise generated by its previous generation cards so they worked on getting the current generation card to run quieter by adopting the same vapor chamber mechanism and fan optimizations used by ATI cards, and while there are lots of reviews claiming the card is silent at idle, I do not believe this is the case per SPCR standards (flames to /dev/null please).
Hardware Labs is a renowned maker of fancy radiators and they market the SR-1 line as being optimized for silent operation. The test subject for this review is an SR-1 120. “120” referring to the size of the cooling area which happens to exactly match a 120x120mm fan. Most people who would attempt to cool a hot beast like the 580 would probably go for a radiator two or three times bigger than a single 120. With that in mind, let us take a look at photos of the box and each unwrapping step as we remove the subject from its container…just kidding. Below is the test subject along with accessories it came with.
The radiator came with 1/2” inner diameter bards and mounting screws to mount the fan to the radiator and the radiator to the case. I replaced the 1/2” ID barbs with 3/8” ID barbs to match the hose size I use.
Below is a photo with the nexus fan to be used in this test installed.
SR-1 has two main features to optimize for low air flow conditions. First is the wider than usual fin spacing on the radiator, second is the gap between the radiator case and the fins which is supposed to help with the airflow between the fan and the radiator fins. Below is a shot of the radiators showing fin spacing.Test Setup and Test Methodology
- NVidia GTX580 at stock clock speeds
- HWLabs Black Ice SR-1 120 radiator, externally mounted (outside the case)
- Nexus 120mm fan connected to a fanmate controller
- Danger Den DD-GTX580 waterblock
- Swiftech MCP350 pump with XSPC DDC acrylic top connected to a variable DC supply
- Intel I7-920
- Power meter to monitor power usage from the wall
- Furmark 1.6.5 and Prime95 to load the system
Below is the GPU waterblock on my GTX580. The block is a solid piece of milled copper. Notice how the card is sloping to one side.
The watercooling loop in my test system does not make use of a reservoir but does use two T-lines and Koolance quick disconnects to help fill the loop with water. Plain distilled water without any additives is used to fill the loop.
Temperature measurements were taken by first heating up the system using furmark with the pumps running at 10v and fans at 7v. Once heated up and stable, the system was cooled down by running the pumps and fan at full speed until system temperatures once again stabilized (~20-30 minutes). Pump and fan voltages were then set at the starting level for the run, successive measurements taken each time fan voltage was lowered and temperature has stabilized. The highest temperature reached after temperature stabilized is recorded. Another cooling down period was initiated prior to the next set of pump-fan voltage measurements.Power Measurements
I have no way to measure power consumption other than the power meter attached to my PC. To estimate the card's power consumption, I took a series of measurements and subtracted out as much of the rest of the system's power consumption that I could. Note too that I am using an older version of furmark which does not appear to be throttled since I didn't see the saw tooth pattern in newer versions other reviewers saw:
Seven Prime95 theads were used to fully load the system because Furmark needs one thread to keep it fed otherwise the GPU is not fed enough work and power consumption goes down.
We can guestimate the GPU only power consumption by doing the following:MaxConsumption
= GPUConsumption * PSUefficiency = approximate GPU power consumption
Substituting the numbers, we get
(565W – 270W + 26W) * .85 = 273W
XBitLabs measured the GTX 580’s power draw in OCCT and came up with 288W so my computations are not far off from their measurements.
Without further ado (and mostly because the length of this post has exceeded my attention span
), here are the temperature measurements. First, stock GTX580 measurements inside an Antec P183 case. Ambient is at 20C.
The folding and furmark measurements are taken with the front door of the P183 open, otherwise the temperatures will go up some more. I stopped the furmark measurement when the GPU temps hit 95C.
Temperature measurements varying pump and fan voltage, ambient at 20C.
With temperatures a little over 40C over ambient on full load, it’s right where my card would be idling on a dual monitor setup. The temperatures start going up faster with the fans at 7V. My guess is at those voltage levels, the fans are not moving enough air to take the heat out fast enough.
What about noise, this is SPCR after all. The quality of the Nexus fans are well known on this site. I do have to mention that on mine, I hear a regular tocking sound when I put my ear next to the fan. As far as the pump, between 9-10V unenclosed, with some measures taken to reduce vibration, it is silent enough at around three/four feet away that I could not hear it over the noise of the hard drives. I will go over its noise characteristics in another review (maybe).
- Keeps GPU cooler compared to stock cooler
- Good VRM cooling (since they are presumably on the same block)
- Quieter than stock cooling solution
- Does not exhaust hot air inside the case (when mounted externally)
- Cost more than third-party air-coolers
- Higher complexity
- More maintenance required
- Greater chance for failure
- GPU block is specific to card model and needs to be replaced when upgrading GPU.
So there you go. Ignoring the other factors such as price, complexity of watercooling systems, “overloaded” loop, the conclusion appears fairly obvious. So now, about those hard drives...